was deleted from the filesystem to the disk driver, commonly
known as "discard" or "trim".
fs/driver support is in ffs and ata wd for now.
This is what was posted here:
http://mail-index.netbsd.org/tech-kern/2012/02/28/msg012813.html
with minor cleanup, and the global switch replaced by a mount option.
and block allocation).
When quota2_check() is called with an uid not yet in the list,
getinoquota2() will call quota2_q2ealloc() to allocate a new entry for this
uid. quota2_q2ealloc() will remove an entry from the free list and
put it at the head of the corresponding hash list, and flush the block
containing the header if it's not the one also containing the allocated entry.
quota2_q2ealloc() then return the alocated entry and corresponding block
to caller (getinoquota2() here), which returns it to quota2_check().
quota2_check() then checks if the allocation can succeed, and returns and
error if not and calls brelse() on the buffer (because from his POW no
change was made to the entry), effectively discarding changes
to the entry that may have been made by quota2_q2ealloc().
Fix by always bwrite()ing the entry in quota2_q2ealloc(), and re-reading
the entry in caller.
Add a bunch of kasserts to check more stringently that ufs_direnter
did not compact across directory blocks. Don't bother fetching
subsequent I/O blocks from the directory: ufs_lookup guarantees that
it's not necessary, and the kasserts check this to be sure.
The message fired when we were looking at the start of an I/O block,
not when we crossed from the end of one to the start of another. I
believe it fired only when tulr->ulr_offset was a multiple of the I/O
block size (fs_bsize), which can happen if ufs_lookup either finds an
entry or finds free space at the start of an I/O block.
If ufs_lookup found an entry, none of this ulr recalculation logic
should kick in -- if tvp != NULL, then tulr->ulr_count is garbage, so
it's not merely unnecessary but wrong (although I suspect harmless in
the end) to read it in ufs_rename_overlap_p in consideration of
whether to recalculate fulr.
Discussed with chuq and dholland.
ok dholland
all we decided to adopt the Linux API, therefore there is rationale
to stick to it.
No standard tells us what to do, and our extended attribute API has not
been used in a release, therefore we do not break anything, and we get
more easily compatible with programs using the Linux extended attribute
API.
Note that FreeBSD and MacOS X return ENOATTR. FreeBSD has its own API
and MacOS X has a Linux-like API. How did the world get so complicated?
translate FSYNC_LAZY into PGO_LAZY for VOP_PUTPAGES() so that
genfs_do_io() can set the appropriate io priority for the I/O.
this is the first part of addressing PR 46325.
Adapt dholland@'s fix to ufs_rename to fix PR kern/43582. Address several
other MP locking issues discovered during the course of investigating the
same problem.
Removed extraneous vn_lock() calls on the Ifile, since the Ifile writes
are controlled by the segment lock.
Fix PR kern/45982 by deemphasizing the estimate of how much metadata
will fill the empty space on disk when the disk is nearly empty
(t_renamerace crates a lot of inode blocks on a tiny empty disk).